[release/2.4] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x #1688

mhalk · 2024-11-11T18:28:47Z

Replace (more) exact calculation with hardware approximation.

Benefits:
Reduced code size.
Improved performance for certain scenarios.

Experiments show low reduction in precision.
Experiments show no significant performance regressions.
bfloat16 as well as float16 related calculations may benefit largely from this change.

vectorized_layer_norm_kernel:
Gains performance esp. for the following tensor shapes.
Lower values for dim1 do not change performance significantly.
dim1 = 8k-65k may gain considerable performance, but decline gradually with size.

dim0    dim1
----    ----
1024	8192
1024	16384
1024	32768
1024	65536
1024	131072
1024	262144
1024	524288

rocm-mici · 2024-11-11T19:37:27Z

Jenkins build for 86e72f59ea71082a578ecaef6717677856675fb4 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

	/lib/x86_64-linux-gnu/libm.so.6
[7979/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_shifted_chebyshev_polynomial_w.hip.o
[7980/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_BinaryMiscOpsKernels.hip.o
[7981/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_spherical_bessel_j0.hip.o
[7982/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/./torch_hip_generated_attention_backward.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/attention_backward.hip:49:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-12T23:18:13Z

Jenkins build for 86e72f59ea71082a578ecaef6717677856675fb4 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

[7980/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_shifted_chebyshev_polynomial_u.hip.o
[7981/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_spherical_bessel_j0.hip.o
[7982/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_ValidateCompressedIndicesKernel.hip.o
[7983/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_UnaryGeometricAcoshKernel.hip.o
[7984/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/./torch_hip_generated_flash_api.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/flash_attn/flash_api.hip:57:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-13T08:40:45Z

Jenkins build for 86e72f59ea71082a578ecaef6717677856675fb4 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

/var/lib/jenkins/pytorch/aten/src/ATen/native/hip/DistributionTemplates.h:186:17: warning: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering [-Wpass-failed=transform-warning]
/var/lib/jenkins/pytorch/aten/src/ATen/native/hip/DistributionTemplates.h:186:17: warning: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering [-Wpass-failed=transform-warning]
/var/lib/jenkins/pytorch/aten/src/ATen/native/hip/DistributionTemplates.h:186:17: warning: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering [-Wpass-failed=transform-warning]
12 warnings generated when compiling for gfx942.
[7983/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/./torch_hip_generated_attention_backward.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/attention_backward.hip:49:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Replace (more) exact calculation with hardware approximation. Benefits: Reduced code size. Improved performance for certain scenarios. Experiments show low reduction in precision. Experiments show no significant performance regressions. bfloat16 as well as float16 related calculations may benefit largely from this change. vectorized_layer_norm_kernel: Gains performance esp. for the following tensor shapes. Lower values for dim1 do not change performance significantly. dim1 = 8k-65k may gain considerable performance, but decline gradually with size. dim0 dim1 ---- ---- 1024 8192 1024 16384 1024 32768 1024 65536 1024 131072 1024 262144 1024 524288 --------- Co-authored-by: Hashem Hashemi <hashem.hashemi@amd.com>

rocm-mici · 2024-11-13T19:44:05Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

  580 | #define DEPRECATED(msg) __attribute__ ((deprecated(msg)))
      |                                         ^
1 warning generated when compiling for gfx908.
[7941/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_scaled_modified_bessel_k0.hip.o
[7942/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/./torch_hip_generated_flash_api.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/flash_attn/flash_api.hip:57:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-13T20:01:05Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

/opt/rocm-6.2.3/lib/llvm/bin/../../../include/hip/hip_runtime_api.h:580:41: note: expanded from macro 'DEPRECATED'
  580 | #define DEPRECATED(msg) __attribute__ ((deprecated(msg)))
      |                                         ^
1 warning generated when compiling for gfx908.
[7944/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/./torch_hip_generated_flash_api.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/flash_attn/flash_api.hip:57:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-14T03:13:03Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

[7938/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adam_impl.hip.o
[7939/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adamw_impl.hip.o
[7940/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_legendre_polynomial_p.hip.o
[7941/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_shifted_chebyshev_polynomial_t.hip.o
[7942/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/./torch_hip_generated_flash_api.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/flash_attn/flash_api.hip:57:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-14T05:37:50Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

rocm-mici · 2024-11-15T05:53:03Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

rocm-mici · 2024-11-15T06:42:38Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

[7937/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adamw_impl.hip.o
[7938/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adamw_amsgrad_impl.hip.o
[7939/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_UnfoldBackwardKernel.hip.o
[7940/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adam_amsgrad_impl.hip.o
[7941/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/./torch_hip_generated_attention_backward.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/attention_backward.hip:49:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-15T19:08:48Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

  580 | #define DEPRECATED(msg) __attribute__ ((deprecated(msg)))
      |                                         ^
1 warning generated when compiling for gfx908.
[7983/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_spherical_bessel_j0.hip.o
[7984/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/./torch_hip_generated_flash_api.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/flash_attn/flash_api.hip:57:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

rocm-mici · 2024-11-19T20:46:05Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

[7941/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_modified_bessel_k0.hip.o
[7942/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adamw_impl.hip.o
[7943/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_scaled_modified_bessel_k1.hip.o
[7944/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_scaled_modified_bessel_k0.hip.o
[7945/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/./torch_hip_generated_attention_backward.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention_backward.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/attention_backward.hip:49:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

okakarpa · 2024-11-21T17:42:17Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

jeffdaily · 2024-11-21T18:29:06Z

How do we fix the CI? Do we want to just try an upstream draft PR so we can get a CI signal?

jerrymannil · 2024-11-21T19:16:56Z

@mhalk

Can you also run pytorch UTs that uses layer norm ?
Code search results

Code search results

After that you can also try to run the full pytorch UT test suite
CONTINUE_THROUGH_ERROR=1 .ci/pytorch/test.sh

pruthvistony · 2024-11-21T21:29:43Z

How do we fix the CI? Do we want to just try an upstream draft PR so we can get a CI signal?

I agree with @jeffdaily suggest.
Since our internal CI is broken, if the __builtin_amdgcn_rcpf is available in ROCm 6.2 we should raise a PR to upstream.

pruthvistony

LGTM.
if no regression on UTs, good to merge.

okakarpa · 2024-11-22T18:33:18Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

[7936/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_hermite_polynomial_h.hip.o
[7937/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adam_impl.hip.o
[7938/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_fused_adam_amsgrad_impl.hip.o
[7939/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/hip/torch_hip_generated_scaled_modified_bessel_k0.hip.o
[7940/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/./torch_hip_generated_flash_api.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/flash_attn/torch_hip_generated_flash_api.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/flash_attn/flash_api.hip:57:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

jerrymannil · 2024-11-22T20:50:00Z

upstream PR paased CI. See pytorch#141309

okakarpa · 2024-11-23T09:42:42Z

Jenkins build for 1cd0f364e5e7681a3f71ff81ceb2b6fd5eefab81 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

mhalk requested review from carlobertolli, jeffdaily, doru1004, pruthvistony, jerrymannil and akadutta November 11, 2024 18:28

mhalk force-pushed the mhalk/feat/layer_norm_reciprocal_builtin branch from 86e72f5 to 1cd0f36 Compare November 13, 2024 19:22

pruthvistony approved these changes Nov 21, 2024

View reviewed changes

jerrymannil changed the title ~~[layer_norm][AMDGPU] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x~~ [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x Nov 22, 2024

jerrymannil approved these changes Nov 22, 2024

View reviewed changes

jithunnair-amd changed the title ~~[ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x~~ [release/2.4] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x Nov 25, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[release/2.4] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x #1688

[release/2.4] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x #1688

mhalk commented Nov 11, 2024 •

edited

Loading

rocm-mici commented Nov 11, 2024

rocm-mici commented Nov 12, 2024

rocm-mici commented Nov 13, 2024

rocm-mici commented Nov 13, 2024

rocm-mici commented Nov 13, 2024

rocm-mici commented Nov 14, 2024

rocm-mici commented Nov 14, 2024

rocm-mici commented Nov 15, 2024

rocm-mici commented Nov 15, 2024

rocm-mici commented Nov 15, 2024

rocm-mici commented Nov 19, 2024

okakarpa commented Nov 21, 2024

jeffdaily commented Nov 21, 2024

jerrymannil commented Nov 21, 2024

pruthvistony commented Nov 21, 2024

pruthvistony left a comment

okakarpa commented Nov 22, 2024

jerrymannil commented Nov 22, 2024

okakarpa commented Nov 23, 2024

[release/2.4] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x #1688

Are you sure you want to change the base?

[release/2.4] [ROCm][layer_norm] Use __builtin_amdgcn_rcpf(x) instead of 1.f/x #1688

Conversation

mhalk commented Nov 11, 2024 • edited Loading

rocm-mici commented Nov 11, 2024

rocm-mici commented Nov 12, 2024

rocm-mici commented Nov 13, 2024

rocm-mici commented Nov 13, 2024

rocm-mici commented Nov 13, 2024

rocm-mici commented Nov 14, 2024

rocm-mici commented Nov 14, 2024

rocm-mici commented Nov 15, 2024

rocm-mici commented Nov 15, 2024

rocm-mici commented Nov 15, 2024

rocm-mici commented Nov 19, 2024

okakarpa commented Nov 21, 2024

jeffdaily commented Nov 21, 2024

jerrymannil commented Nov 21, 2024

pruthvistony commented Nov 21, 2024

pruthvistony left a comment

Choose a reason for hiding this comment

okakarpa commented Nov 22, 2024

jerrymannil commented Nov 22, 2024

okakarpa commented Nov 23, 2024

mhalk commented Nov 11, 2024 •

edited

Loading